Table of Contents
Journal Editor & Reviewers
PREDICTING RANK ORDER STOCK PRICE PERFORMANCE USING A MULTI-FACTOR RELATIVE PRICE STRENGTH MODEL Frederic H. Dickson, CMT
SCIENCE IS VALIDATING THE CONCEPT OF THE WAVE PRINCIPLE Robert R. Prechter, Jr., CMT
THE INTERACTION OF TRENDINESS MEASURES AND TECHNICAL INDICATORS Basil Panas, CFA, CPA, CMT
HEAD AND SHOULDERS ACCURACIES AND HOW TO TRADE THEM Serge Laedermann
VOLATILITY AND STRUCTURE: BUILDING BLOCKS OF CLASSICAL CHART PATTERN ANALYSIS Daniel L. Chesler, CTA, CMT
Journal Editor & Reviewers
Henry 0. Pruden, Ph.D. Golden Gate University San Francisco, Calijinnia
David L. Upshaw, CFA, CMT Lake Quivira, Kansas
Jeffrey Morton, M.D. PRISM Trading Advisors Missouri City, Texas
Connie Brown, CMT Aerodynamic Investments Inc. Pawlq’s Island, South Carolina
John A. Carder, CMT Topline Investment Graphics Boulder, Colorado
Ann F. Cody, CFA Hilliard Lyons Louisville, Kentucky
Robert B. Peirce Cookson, Peirce & Co., Inc. Pittsburgh, Penns$vania
Charles D. Kirkpatrick, II, CMT Kirkpatrick and Company, Inc. Chatham, Massachusetts
John McGinley, CMT Technical Trends Wilton, Connecticut
Cornelius Luca Bridge Information Systems New Ymk, New Yorlz
Theodore E. Loud, CMT Tel Advisor Inc. of Virginia Charlottesville, Virginia
Michael J. Moody, CMT Dory, Wtigh t & Associates Pasadena, Calijornia
Richard C. Orr, Ph.D. ROME Partners Marbtehead, Massachusetts
Kenneth G. Tower, CMT UST Securities Princeton, New Jersey
J. Adrian Trezise, M. App. SC. (II) Consultant toJ.P Morgan London, England
Barbara I. Gomperts Financial & Investment Graphic Design Marblehead, Massachusetts
Market Technicians Association, Inc. One World Trade Center; Suite 4447 New Ymk, New Ymk 10048
Return to Table of Contents
1: PREDICTING RANK ORDER STOCK PRICE PERFORMANCE USING A MULTI-FACTOR RELATIVE PRICE STRENGTH MODEL
Frederic H. Dickson, CMT
One of the greatest challenges facing equity investors is predicting individual stocks’ relative future price performance in a manner that is disciplined, can be easily replicated and produces consistently accurate results over the investment time horizon of interest to the user.
As Research Director of a regional brokerage firm, I am continually asked to offer an opinion regarding the future price performance of specific stocks relative to a specific universe or specific portfolio. To meet this challenge, I have constructed and currently maintain and electronically distribute an extensive equity database. Updated weekly, this database includes a wide variety of technical and fundamental indicators and several forecasting models, including a short-term, technically-based, relative strength momentum model designed to provide relative performance guidance over a three to six-month time horizon.
Currently there is no shortage of proprietary and publicly available research tools attempting to accomplish this objective. In the author’s opinion, there does exist a noticeable absence of published data evaluating how well these tools work “after-the-fact,” assuming multiple start and end dates. We continue to observe that most published test results are derived from back-testing procedures assuming a single start and end date for the test. A potential user of the indicator is often at a loss to determine if encouraging results are the product of a model that has consistent forecasting ability or merely a coincidentally favorable test period.
Finally, the prospective user of a forecasting or relative ranking system often has no idea of how long the projected rankings will provide predictive value before deteriorating, or the consistency of a ranking methodology in accurately predicting the rank order of investment results for a significant equity universe under consideration. Results are often reported from universe subsets that will provide encouraging results. In summary, we have typically found the absence of data on the pervasiveness and consistency of test results generated by systems employed live or “after-the-fact” to be very troubling.
OVERVIEW AND CONCLUSIONS
This paper describes and presents the results of a technically based, multifactor stock selection and ranking index (momentum index) research methodology. The results presented are based on an ex post facto analysis of actual predicted and published rank performance suggested by the index. The rankings have been published weekly as part of the Branch Cabell Equity Advantage Database since June 25, 1999. The test analysis extends from the initial index publication date ofJune 25,1999 through November 26, 1999. The testing protocol considers the performance of the index assuming multiple overlapping start and end dates (of variable length holding periods) during this time period. As described below, the initial test results are encouraging, as the model appears to have provided positive predictive value over a wide variety of holding periods as determined using several rigorous academically acceptable evaluation criteria.
The momentum index in Chart 1 below shows the rank of an individual stock relative to its selection universe based on combining two ranked measures of cycle position for each stock and three ranked measures of price change.
Our investment hypothesis is that the Branch Cabell momentum index can demonstrate consistent predictive rank order performance results in excess of those generated from investing in a market (S&P 500) index fund over various weekly holding periods after making appropriate adjustments for historical price risk.
The momentum index was initially created to help investors assess probable three to six month rank order price for a 1,750 com pany equity universe including approximately 100 listed ADRs. The debut of this indicator was June 25,1999. As shown in Chart 1, the S&P 500 experienced two corrections and recoveries of at least five percent between July 1999 and November 26,1999. Looking back, this introductory five-month period was extremely trying for most investors as well as being a very diflicult period to test and evaluate any technically-based stock selection methodology due to the number and magnitude of the market and individual stock price directional changes. Over the entire time period, the S&P 500 was up 6.5% and the average price change of stocks included in the test universe was down 2%:
For testing purposes, we assumed an equal dollar-weighted investment in each name in the universe each week. We then divided the universe into deciles based on the ranking suggested by the momentum index and measured the performance of each ranking decile weekly over various holding periods during the test period. This procedure eliminated the possibility of favorable start and end dates impacting the test results. Although five months of test data is a very short time frame to evaluate an index and to conclude the validity of our investment hypothesis, we believe the following initial observations are noteworthy and justify continued publication of the index and the indefinite extension of the testing procedure.
The momentum index successfully projected rank-order performance over 20 overlapping time periods (extending from l-20 weeks) based on multiple start and end dates evaluated between June 25, 1999 and November 26, 1999 (Table 1). The results were pervasive and surprisingly consistent over the range of multi-week holding periods. The correlation coefficients of the momentum index ranked order performance ranged between 0.75 and 0.89 (1.0 marks perfect correlation, 0.0 marks zero correlation and -1.0 marks perfect negative correlation) for all time periods tested.
The absolute returns also initially suggest a high degree of rankorder forecasting ability. For all periods tested, the absolute return of stocks ranked in the top decile ranking was greater than the returns produced by stocks in the second decile (See Chart 3). Stocks ranked in the 2nd decile in turn outperformed stocks ranked in the 5th decile then in turn outperformed stocks ranked in the last (10th) decile. The degree of outperformance between the stocks in the top ranked decile and bottom ranked decile ranged from an average of 2.3% for a l-week holding period to 19.4% for a IO-week holding period and 40.3% for a 20-week holding period. The spread of rank order returns are highly significant when compared to the distribution of returns generated from a random selection of stocks made from the same universe tested in a similar manner over the same time period (Table 2). The top decile of stocks selected randomly underperformed the bottom ranked decile by 0.3% for one week, out erformed by 0.08% at 10 weeks and underperformed by 1.17 o at 20 weeks. ?
Stocks identified in the top two ranked deciles produced positive risk-adjusted excess returns for all time periods up to 17 weeks when evaluated using the Jensen Modified Capital Asset Pricing model (Table 3). The results were dramatically above what a rational investor would expect based on the risk profile of the stocks included in each decile category. Stocks included in the bottom two ranked deciles consistently produced the poorest negative excess returns over the entire spectrum of holding period.
The momentum rankings index produced excess returns consistent with their decile position rather than the average beta associated with each decile ranking position. These results were inconsistent with what one would expect based on the volatility assigned to each decile ranking class based on historical betas. This apparent market anomaly is worth noting and strongly suggests that future tests be conducted to determine the extent and pervasiveness of this anomaly over longer time periods including a full market and economic cycle.
We expected the average volatility, as measured by beta, for the stocks in each momentum index decile to decline proportionately by decile ranking category. We expected the highest momentum index ranked stocks to have the highest average historical beta and the lowest ranked stocks to have the lowest historical beta. In fact, the observed average beta declines sequentially as expected between decile ranks 1 and 5 but then unexpectedly rises sequentially between decile ranks 6 and 10 (Table 3).
The average beta measured over the entire 20-week time horizon within specific momentum ranking deciles was not stable (Table 4). During one period of sustained market weakness, (July 16July 30) the average beta of the top decile momentum ranked stocks fell from 1.38 to 1.15 while the beta of the lowest momentum ranked stocks rose from 1.04 to 1.20. The average beta of the middle-ranked decile remained very stable throughout the entire test period. The unusual variability could possibly be attributed to stocks eliminated from the universe during the testing period that were replaced by stocks with substantially different volatility characteristics.
We expected the momentum index to demonstrate proportionately reduced forecasting ability as the holding period lengthened. The test data suggests that the momentum index’s ability to produce returns consistent with the rankings persists much longer than we originally expected. Although we have only a few data points for holding periods beyond 15 weeks, the rank order correlation coefficients remain very high (0.80) with little noticeable deterioration beyond this time horizon. The positive spread of realized returns between performance ranks remains intact from the highest decile to the lowest decile for all periods up to 20 weeks. For this limited testing period, the momentum index met our initial objective of pervasiveness by maintaining its discrimination ability across the stock universe for time periods in excess of 13 weeks.
We observed significant deviation of returns for the individual stocks included within each of the decile rankings. The performance statistics of individual stocks suggest the widest dispersion of individual stock returns at the highest and lowest decile ranking levels. Therefore, one needs to look at the decile performance rankings as only an indication of central tendency for the stocks included in each decile rather than an absolute predictor of future individual stock performance. The performance ranks suggest probability of performance rather than serving as an explicit predictor of performance on a stock-by-stock basis.
We conclude that for the time period tested, the momentum index provided valuable forecasting information about the future risk-adjusted excess returns that could be profitably exploited by investors after considering reasonable transaction costs. An investor could have begun to employ the published momentum index rankings several weeks after the testing period began and would have received approximately the same benefit as an investor who employed the model from the start of the test period over the entire array of holding periods. The results appear to be consistent and pervasive during the test period across holding periods ranging from one to twenty weeks.
METHODOLOGY The Momentum Ranking Index
Background. The genesis of the author’s interest in relative strength analysis dates back over 30 years. In his 1967 doctoral thesis, Dr. Robert A. Levy scientifically explored and tested a 26 week relative strength ranking system that he claimed invalidated the widely accepted “weak efftcient market thesis.” Several academic researchers at the time concluded that Dr. Levy’s ability to demonstrate exceptional performance results was a direct function of the volatility inherent in the stocks selected rather than a persistent market anomaly. Thus, Dr. Levy’s claim of refuting the efficient market hypothesis was widely discredited. On a practical basis, we have found the original 26week rate of change indicator to be helpful in establishing probabilities of future results, but lacking persistence and consistency when applied across a wide universe of stocks.
Index definition and construction. The momentum ranking index is constructed using only historical price behavior of individual stocks. Thus, it is a pure “technical” index. Conceptually, the index attempts to quantify a stock’s position within a 52-week price cycle and its momentum or rate of change as measured over 4week, 13-week and 52-week periods. The momentum ranking index subcomponents, cycle position and velocity (percent price change) appear to be greatly impacted by overall market factors. The ability of the stock to respond to changing market factors is hypothesized to be a critical variable in determining near-term price changes. This index has been continuously constructed on a weekly basis since June 25,1999. No changes were made in construction methodology during the test period.
Each week every stock in the 1,750 company universe is ranked relative to the entire universe based on its respective Price/52-week high and Price/52-week low to determine relative cycle position. Then each stock is separately ranked on the basis of its 4week, 13 week and 26week price change relative to the same universe. Each stock’s ranked position based on each of these five criteria are then summed and ranked relative to each stock in the universe to determine the final technical momentum ranking index. A stock ranking number 1 in each category would have a composite score of 5. This score would be compared to the scores of all other companies in the universe to determine a final momentum index rank. The stock with the lowest cross-ranked score is projected to have the highest probability of outperforming all other stocks in the universe going forward (See Chart 1).
During the testing period, approximately 75 companies from the original starting universe were eliminated from the universe due to mergers or acquisitions. New companies were introduced into the universe during the test period at the request of our retail clients, our institutional brokerage clients or to include IPOs of technical or fundamental interest when data became available on the StockVal database. For companies with less than 52 weeks of pricing data, we calculated comparable cycle position statistics using Price/Life of Company high price in place of the Price/52 week high ratio and Price/Life of Company low price in place of the Price/52-week low ratio. For companies with fewer than 13 weeks of pricing data, we substituted the price change from the company’s IPO to the calculation date for the index in the velocity indicators. We have not identified the impact of these changes on the test results shown in this paper.
The momentum index is calculated based on Friday closing prices (4:30 PM EST/EDT) and does not recognize prices posted in Friday aftermarket trading on electronic exchanges such as Instinet. The historical prices in the database are adjusted when a stock split or meaningful stock dividend occurs. Companies that have been acquired during the test period are purged from the universe to preserve comparability of companies from each weekly starting point. This adjustment might add a small positive or negative bias to the test results.
Test period. The test period was conducted between June 25, 1999 and November 26, 1999 using the technical momentum index published weekly in the Branch Cabell Equity Advantage Database between June 25,1999 and November 5,1999. June 25,1999 marked the first date the Technical Momentum Index was pub lished and distributed to clients.
Stock Universe. The Equity Advantage Stock universe was originally constructed in October 1998. It includes members of the S&P 500, the Russell 1000, selected holdings or stocks of special interest to clients of Branch Cabell, and stocks covered by CS First Boston and Prudential Research (research correspondents of Branch Cabell). Stocks not otherwise identified with at least $1 billion in market capitalization are also included in the database. The performance of the Branch Cabell Equity Universe versus the S&P 500 is shown in Chart 2. The stocks included in the universe are included in the StockVal” database which is used as the basic information source for all data. Friday night closing prices are downloaded from the StockVal” database and loaded into the Branch Cabell Equity Advantage database every Saturday. StockVal’” provides component calculations for the five variables included in the Technical Momentum Index.
Testing Protocol. Each week the technical momentum ranks and individual equity betas were loaded into an Excel spreadsheet along with the model ranking algorithms. Historical weekly prices were retrieved from the StockVal” database for each worksheet, providing the necessary data to calculate cumulative weekly returns from the initial date of the holding period to the last date included in the test (November 26, 1999). The stock prices were split-adjusted but were not adjusted for spinoffs that may have negatively impacted the performance of a specific stock. Each weekly database was then sorted in ascending order of technical momentum rank, with most favorable momentum rank at the top of the list and least favorable at the bottom of the list. The universe was then divided into deciles, and average performance returns were calculated for each performance decile. The data were ordered so that the average performance of comparable weekly holding periods could be determined. The procedure was repeated for each of the twenty weeks included in the test. The results were averaged for each ranking decile by comparable holding periods. Thus one could easily evaluate the returns for all l-week, 5-week, lo-week, etc. holding periods on a common basis.
This procedure allows us to draw conclusions about the persistence and consistency of the performance ranking results without assuming specific starting and ending test period dates. We view this as a very rigorous but fair testing protocol. The results of this protocol are shown in Table 1. Chart 3 presents a graph of the test results over the test period. After 20 weeks, initial signs of conver gence between the performance of the bottom decile and the middle decile ranking position were beginning to appear, although the number of data points observed remain very small (3). The spread between the top decile ranking position and the middle decile ranking position continued to widen.
Mindful of the “weak efficient market hypothesis” which suggests that purely historical stock price behavior has no predictive power, we decided to construct a benchmark test assigning random numbers as a pseudo technical momentum rank, or “pseudo ranks.” Using the Excel worksheet’s random number function, a number between 0 and 1 was generated and multiplied by the universe size to determine a stock’s pseudo rank. Stock performance tests were then conducted in a manner consistent with the test procedure used to determine the performance of the technical momentum ranks. The data from this test is shown in Table 2. The randomly generated performance ranks produced apparently random results within very tight performance boundaries. The re sults of the “pseudo ranking” test provide a benchmark in order to evaluate whether our technical momentum model was the product of a random process or identified a market anomaly that can be exploited by investors. Performance that substantially exceeded the randomly generated results, particularly at the decile rank ex tremes, added confidence in the validity of the momentum index test results.
A comparison of the performance of the technical momentum ranks versus the “pseudo ranks” strongly suggests that the predictive performance of the technical momentum rank was the result of a process other than chance. We draw the same conclusion evaluating the average rank order correlation coefftcients of the technical momentum ranks (consistently above 0.75 with 99% of the observed individual cell rankings above 0.1) versus the correlation coefficients produced by the “pseudo ranks.” As expected and shown in Chart 3, the performance spread between the decile rankings for the “pseudo ranks” was very narrow and the decile performance showed a high tendency for convergence.
Cognizant of the academic arguments raised in the challenge of Dr. Levy’s study, we then constructed a matrix that identified the betas associated with the stocks grouped into the decile categories by their technical momentum rank. Table Four presents this data. The betas shown were calculated as of September 30, 1999. It was not practical to recreate the betas for June 25, 1999. Our assumption is that the change in betas on a stock-by-stock basis would be minor, as the beta calculation was made based on five years of weekly price data for each stock and for the S&P 500.
The data provided an interesting twist. We expected to see rank order correlation between the betas for each decile and the momentum index decile rankings. This would indicate that the stocks with the highest estimated technical momentum would have the highest betas and those with the lowest technical momentum would have the lowest betas. The data did not confirm this hypothesis. In fact, the data suggest a bi-modal distribution with the betas accelerating as one approaches the upper and lower decile ranking levels. We did not expect the worst performers to have the second highest decile beta rankings in the universe during the test period.
As a final test, we decided to compare the performance results produced by the technical momentum rankings to those predicted by the Jensen Modified Capital Asset Pricing Model (MCPM), a benchmark test used to determine rational asset pricing. MCPM states that an asset’s return is related to the risk free rate of return plus the difference between market rate of return (S&P 500) and the risk free rate of return times the beta of the specific security. (Expected Individual Security Return = Risk Free Rate t (Market Return - Risk Free Rate)* Individual Security Beta). If the differential is positive, an unexplained “excess return” is generated. Investors are being compensated for their unusual investment knowledge.
Table 3 presents the excess returns generated using the momentum rankings by decile over the test period, assuming various holding periods and starting dates. The theory behind the MCPM assumes that the return of the asset category will be a direct function of the asset category’s volatility as measured by beta. The data shown below contradict that conclusion. The excess returns systematically decreased in direct proportion to the rank ordered position of the index in contradiction to the directional movement of the average beta by decile position. This anomaly is certainly worth exploring in more depth in the future as the momentum index gains more ex post facto history.
Our hunch is that the anomaly partially reflects the fact that the measurement period of the performance data is far shorter than the time period used to calculate each individual stock’s beta. We believe betas calculated for a time period consistent in length with the test period could have produced far different and more predictable results consistent with that expected using the MCPM. Thus, we cannot make a strong assertion about the validity of the Capital Asset Pricing Model when evaluated from the perspective of this test protocol. The data do suggest that the technical price momentum model successfully discriminated future price performance on a rank-order risk-adjusted basis during the test period.
The findings of this study are highly encouraging. The results suggest that momentum as a market behavior force was much more pervasive than we previously expected. Clearly, this is an investment style employed by enough participants in the market place to impact security pricing behavior. We will continue to capture, test and evaluate future results using the ability of the momentum index rankings to predict rank order stock performance behavior over varying time horizons. In the future, we plan to evaluate the performance of the technical momentum performance ranks on the basis of market capitalization to determine if there is any small or large cap bias and in combination with our fundamentally based indicators. Our goal is to understand how well our published indicators work, why they work, to identify forecasting problems if and when they occur and to encourage other practicing technical analysts to adapt a similar rigorous approach to testing the validity of their model forecast on an ex-post-facto basis.
Robert A. Levy, “Random Walks, Realty or Myth,” Financial AnalystsJouma1 (November-December 1967a).
Michael C. Jensen and George A. Bennington, “Random Walks and Technical Theories: Some Additional Evidence,” The Journal ofFinance, XXV, No. 2 (May 1970).
Frederic H. Dickson, CMT is Managing Director of Research at Branch Cabell & Co., Inc., in Richmond, VA. Fred is a past President of the Market Technicians Association (1983 1984), served for many years as the Educational Committee Chairman of the MTA and authored the first set of test questions selected for use in the CMT Level I examination. Fred has served as an Adjunct Assistant Professor of Finance at the University of Richmond and as an Instructor at the New York Institute of Finance. He has contributed several articles in the past to the MTA Journal. He presently publishes a daily and weekly market comment and the Branch Cabell Equity Advantage Database for an institutional audience.
Return to Table of Contents
2: SCIENCE IS VALIDATING THE CONCEPT OF THE WAVE PRINCIPLE
Robert R. Prechter, Jr., CMT
New discoveries in the field of complexity theory, fractal geometry, biology and psychology are rapidly yielding more knowledge bolstering the probability that the Wave Principle is a correct description of financial and social reality. This report provides a cursory overview of some of these advances.
To understand the connection between today’s scientific discoveries and the Wave Principle, it is necessary to describe it in modern terms. In the 1930s Ralph Nelson Elliott (1871-1948), through extensive empirical observation, discovered that price changes in stock market indexes produce a limited number of definable patterns (called “waves”) that are variably self-affine’ at different degrees, or sizes, of trend. As opposed to self-identical fractals, whose parts are precisely the same as the whole except for size (see example in Figure l), and indefinite fractals, which are self-similar only in that they are similarly irregular at all scales (see example in Figure 2), Elliott proposed a model of intermediate specificity. Though variable, its component forms, within a defined latitude, are replicas of the larger forms. Waves have event-specific relutiue quantitative properties, as do self-identical fractals, but they are unrestricted in absolute quantitative terms, like indefinite fractals. The fact that both waves and (as we shall soon see) natural branching systems are fractals of intermediate specificity implies that nature uses this fractal style to pattern systems that require highly adap tive variability in order to flourish. Therefore, I think the best term for this variety of fractal is robust fractal. As we shall see, this is a form that living structures typically display.
The essential form of the Wave Principle is five waves generating net progress in the direction of the one larger trend followed by three waves generating net regress against it, producing a three-steps-forward, hvo-steps-back form of net progress. The 5-3 pattern is the minimum requirement for, and therefore the most efficient method of, achieving both fluctuation and progress in linear movement.
Elliott described how waves at each degree become the components of waves of the next higher degree, and so on, producing a structured progression, as illustrated in Figure 3. The word “degree” has a specific meaning and does not mean “scale.” Component waves vary in size, but it always takes a certain number of them to create a wave of the next higher degree. Thus, each degree is identifiable in terms of its relationship to higher and lower degrees This is unlike the infinite scaling relating to clouds or seacoasts and unlike the discrete scale invariance? of simple fractals created by recursive interpolation such as the snowflake in Figure 1. By incorporating features of both, Elliott described a third type of fractal, which we will shortly explore.
Benoit Mandelbrot, an IBM researcher and former professor at Harvard, Yale and the Einstein College of Medicine, did pioneering work bringing to light the fact that fractals are everywhere in nature.3 The term “nature” in this context includes the activities of man, as Mandelbrot began by studying cotton prices’ and most recently presented a multifractal model of the stock market.’ This excerpt from a 1985 article in The Neu York i%res summarizes his exposition on the subject of financial fractals:
Daily fluctuations are treated [by economists] one way, while the great changes that bring prosperity or depression are
This is also what R.N. Elliott said about the stock market sixty years ago. Some members of the scientific community have recently recognized the connection. Three physicists researched the stock market’s log-periodic structures and concluded that R.N. Elliott’s model of financial behavior fits their findings. In 1996, France’s Journal of Physics published the study, “Stock Market Crashes, Precursors and Replicas” by Didier Sornette and Anders Johansen, then of the Laboratoire de Physique de la Matiere Condensee at the University of Nice, France, and collaborator Jean-Phillippe Bouchaud. The authors make this statement:
It is intriguing that the log-periodic structures documented here bear some similarity with the “Elliott waves” of technical analysis [citation EZliott WavePti’ncipk Frost & Prechter] Technical analysis in finance can be broadly defined as the study of financial markets, mainly using graphs of stock prices as a function of time, in the goal of predicting future trends. A lot of effort has been developed in finance both by academic and trading institutions and more recently by physicists (using some of their statistical tools developed to deal with complex times series) to analyze past data to get information on the future. The “Elliott wave” technique is probably the most famous in this field. We speculate that the “Elliott waves” . . . could be a signature of an underlying critical structure of the stock market.
Mandelbrot’s work supports this conclusion. For example, every aspect of Mandelbrot’s general model, as presented in Scientific Am&an,* fits Elliott’s specific model, and no aspect of Mandelbrot’s general model contradicts Elliott’s specific model. Mandelbrot’s work in this regard should properly be seen as compatible with, and therefore support for, Elliott’s more comprehensive hypothesis of financial market behavior. We must also concede the possibility that Elliott’s specific model will be proven false and that financial markets will ultimately be shown to be indefinite fractals, which is as far as Mandelbrot’s work goes. At minimum, though, it may be said that Mandelbrot’s studies are among a number of modem discoveries that increase the probability that RN. Elliott’s fractal model of financial markets is true.
A year after this study (one hopes that it was not in response to it), Mandelbrot published a brief dismissal of Elliott and his work, deriding his predecessor and taking credit for modeling the stock market as a multifractal. (See “Prechter’s Response of Mandelbrot’s Dismissal of Elliott” at www.elliottwave.com/response.htm) Advocates of the Wave Principle are not particularly interested in this controversy per se but in the far more important fact that a renowned scientist has decided that at least one implication of Elliott’s work is so impwtant that he wants creditfm it. Whether that credit is to be taken properly or otherwise is a question for the scientific community to decide, but the key point is that this very situation is yet another fact that increases the potential validity of the Wave Principle hypothesis.
THE ROBUST FRACTAL
It is imperative to understand that R.N. Elliott went fur beyond the comparatively simple idea that financial prices form an indefinite multifractal. One of his big achievements was discovering specific component patterns within the overall form.g Until very recently, it has been generally presumed that there are two types of self-similar forms in nature: (1) self-identical fractals, whose parts are precisely the same as the whole, and (2) indefinitefractalr, which are self-similar only in that they are similarly irregular at all scales. (See Figures 1 and 2.) The literature on natural fractals concludes that nature most commonly produces indefinite fractal forms that are orderly only in the extent of their discontinuity at different scales and otherwise disorderly. Scientific descriptions of natural fractals detail no specific patterns composing such forms. Seacoasts are just Yjagged lines,” trees are composed simply of “branches.” rivers but meander, and heartbeats and earthquakes are merely “events” that differ in frequency. Likewise, financial markets are considered to be self-similarly discontinuous in the relative sizes and frequencies of trend reversals yet otherwise randomly patterned. These conclusions may be due to a shortfall in empirical study rather than a scientific fact.
R.N. Elliott described for financial markets a third type of self-similarity. By meticulously studying the natural world of social man in the form of graphs of stock market prices, Elliott found that there are specific patterns to the stock market fractal that are nevertheless high3 variable within a certain definable latitude. In other words, some aspects of their form are constant and others are vatiabG If this is true, then financial markets, and by extension, social systems in general, are not vague, indefinite fractals. Camp+ nent patterns do not simply display discontinuity similar to that of larger patterns, but th f&m, with a certain latitude, r@icas of them. Elliott defined waves in terms of those aspects that make them identical, thereby allowing for their variability in other aspects of detail within the scope of those definitions. He was even able to define some of the patterns’ variable characteristics in probabilistic terms.
Elliott’s discovery of degrees in pattern formation, i.e., that a certain number of waves of one degree are required to make up a wave of the next higher degree, is vitally important because it links the building-block property of self-identicalfractals to the U’ave Principle, revealing an aspect of self-identity among waves that indefinite fractals do not possess.
Elliott’s discovery of specific hierarchical patterning in the stock market is fundamental. Fractality alone is only a vague comment about that form. Zfpou can describe the pattern, you haue the essence of the object. The more meticulously you can describe the pattern, the closer you get to knowing what it is.
Although Elliott came to his conclusions fifty years before the new science of fractals blossomed, the very idea that financial markets comprise specific forms and identical (within the scope of their definitions) component forms remains a revolutionary observation because, to this day, it has eluded other financial market researchers and chaos scientists. Elliott’s work shows that the general relationship between sizes and frequencies of financial movements, currently considered a breakthrough discovery, is not the essence, but a by-product, of the fundamentals of financial market patterns.
A group of scientists (see below) has very recently recognized that there is a type of fractal in nature whose self-similarity is intermediate between identical and indefinite. As far as I know, theirs is the only published study on the subject. Before we discuss this new aspect of Wave Principle validation, we first must detour through another of R.N. Elliott’s discoveries and understand how it contributes to his grand hypothesis.
THE ROLE OF FIBONACCI IN ROBUST FRACTAIS
Because the essential form of the wave Principle’s is a repeated 5-3, the numbers of waves at different degrees reflect the Fibonacci sequence. The Fibonacci sequence is 1, 1, 2, 3,5, 8, 13, 21, 34, 55, and so on. It begins with the number 1, and each new term from there is the sum of the previous two. The limit ratio between the terms is .618034..., an irrational number sometimes called the “golden mean” but in this century more succinctly phi (4).
The simplest expression of a falling wave is 1 straight-line decline. The simplest expression of a rising wave is 1 straight-line advance. A complete cycle is 2 lines. At the next degree of complexity, the corresponding numbers are 3,5 and 8 (see Figure 4). This Fibonacci sequence continues to infinity.
Both the Fibonacci sequence and the Fibonacci ratio appear ubiquitously in natural forms ranging from the geometry of the DNA molecule to the physiology of plants and animals. Figures 5 and 6 show examples. (For more, see Chapters 3 and 11 in The Wave Principle of Human Social Behavior.) In the past few years, science has taken a quantum leap in knowledge concerning the universal appearance and fundamental importance of Fibonacci mathematics to nature. U’ithout the benefit of that knowledge, after researching the subject to the small extent possible at the time, Elliott presented the final unifying conclusion of his theory in 1940,“’ explaining that the progress of waves is governed by a mathematical principle that governs so many phenomena of life. From this ob servation, he concluded that the progress of mankind is the same type of growth process that we see in so many instances throughout nature.
Modern science is catching up to R.N. Elliott. In 1993, five scientists from the Centre de Recherche Paul Pascal and the Ecole Normale Supeieure in France investigated the diffusion-limited aggregation (DLA) model, which is a set that diffuses via smaller and smaller branches, just like the branching fractals found in nature, such as the circulatory system, bronchial system and trees. Arneodo et al. state at the outset that it is “an open question whether or not some structural order is hidden in the apparently disodered DLA morphology.“” To investigate the question, they use a wavelet transform microscope to examine “the intricate fractal geometry of large-mass off-latice DLA clusers. (See Figure 7)
What mathematics govern this robust fractal? In the first linking (as far as I can discover) of the two concepts of fractals and Fibonacci since Elliott, they demonstrate that their research “reveals the existence of Fibonacci sequences in the internal ‘extinct’ region of these clusters.” The authors find that the branching characteristics of offlattice DLA clusters “proceed according to the Fibonacci recursion law,” i.e., they branch in intervals to produce a l-2-35-8-13-etc. pro gression in the number of branches. The authors of this study, then, have found the Fibonacci sequence in DLA clusters in the samplace that RN. Elliott found the Fibonacci sequence in the Wave Princi$tz in the increasing numbers of subdivisions as the phenomenon progresses.
The authors find even more evidence of Fibonacci. They have discovered that the most commonly occurring “screening angle” between bifurcating branches of these DLA clusters is 36 degrees, which holds regardless of scale. (See Figure 8.) This is the ruling angle of geometric phenomena that display Fibonacci properties, from the five-pointed star (Figure 9) to Penrose tiles (Figure lo), a robust filling of plane-space with just two rhombi. The authors elaborate:
The intimate relationship between regular pentagons and Fibonacci numbers and the golden mean 4 = 2cos(x/5) = 1.618... has been well known for a long time. The proportions of a pentagon approximate the proportions between adjacent Fibonacci numbers; the higher the numbers are, the more exact the approximation to the golden mean becomes. The angle defined by the sides of the star and the regular pentagons is 6 = 36”, while the ratio of their length is a Fibonacci ratio (F,+l/F,).
The authors conclude, “The existence of this symmetry at all sca2RF is likely to be a clue to a structural hierarchical fractal orderng.” Indeed, it is. In a similar way, Elliott found that the price lengths of certain waves are often related by .618, at all scales, revealing another, though perhaps less fundamental, Fibonacci aspect of waves.
These mathematics pertain to “apparently randomly branched fractals that bear a striking resemblance to the tenuous tree-like structures observed in viscous fingering, electrodeposition, bacterial growth and neuronal growth,” which are “strikinglv similar to trees, root systems, algae, blood vessels and the bronchial architecture,” i.e., the typical products of nature.
This is exciting news, but it concerns a model that looks like nature. What do we find when we investigate the actual products of nature? We find phi again and again. In the early 196Os, Drs. E.R. Weibel and D.M. Gomez meticulously measured the architecture of the lung (see Figure 11) and reported that the mean ratio of short to long tube lengths for the fifth through seventh generations of the bronchial tree is 0.62, the Fibonacci ratio.‘* Bruce West and Ary Goldberger have found that the diameters of the first seven generations of the bronchial tubes in the lung decrease in Fibonacci proportionn Oxford professor of mathematics Roger Penrose, who shared the Wolf Prize for Physics in 1988 with cosmologist Stephen Hawking, presents this discussion of the smallest components of our nervous system in his 1994 book, Shadows of the Mind:
The organization of mammalian microtubules is interesting from a mathematical point of view. . ..the skew hexagonal pattern... is made up of 5 right-handed and 8 left-handed helical arrangements... The number 13 features here in its role as the sum: 5 t 8. It is curious, also, that the double microtubules that frequently occur seem normally to have a total of 21 columns of tubulin dimers forming the out side boundary of the composite tube - the next Fibonacci number!“’ [See Figures 12 and 13.1
Led by Eugene Stanley of Boston University, fifteen researchers from MIT, Harvard and elsewhere recently studied the physiology of neurons (see Figure 14) in the central nervous system with the goal of quantifying the arboration of the neurites, which are the arba of neurons. Taking the ganglion cells of a cat’s retina as a model system, they find that the fractal dimension of the cells is “1.68-t or- 0.15using the box counting method and 1.66-t or- 0.08 using the correlation method. “15 Although the authors do not men- tion it, this is quite close to phi. The source of all these biological structures is DNA. Given current best measurements, the length of one DNA cycle is 34 angstroms, and its height is 20 angstroms, very (source: http://polymm bu. edu/)
Recall that each pattern under the Wave Principle has identifiable rigidities as well as tendencies. This is true not only of Elliott waves but of nature’s branching patterns. While the general assumption has been that branching patterns are indefinite fractals, this study shows that these apparently random fractals are in fact more orderly than previouslyrealized. Indeed, Arneodo, et al. determine that they are working with a type of fractal that scientists had not yet found, an intermediate form between exact self-identity and vague, indefinite self-similarity:
The intimate relationship between regular pentagons and Fibonacci numbers and the golden mean...has been well known for a long time.... The recent discovery of “quasicrystals” in solid state physics is a spectacular manifestation of this relationship. This new organization of atoms in solids, intermediate between perfect order and di.sor&, generalizes to the crystalline “forbidden” symmetries, the properties of incommensurate structures. Similarly, there is room for quasifractals between the well-ordered fractal hierarchy of snowflakes and the disordered structure of chaotic or random aggregates. ”
This is the same type of intermediately ordered fractal that R.N.Elliott described for the stock market. I conclude from these studies and the Wave Principle that fractals that characterize nature’s life forms share at least two properties: robustness (inter mediate orderliness/variability) and Fibonacci. I prefer the term robust fractal to quasi-fractal, as its connection to natural, usually living, phenomena indicates that there is nothing quasi about it. I believe that robustness will prove to be the essence of fractals that matter most in nature.”
The latest scientific research is racing headlong toward validating the concept of the Wave Principle, and not just in its simple expression as a financial multifractal. It is also supporting its grander implications that nature’s living fractals are robust, that they are governed by Fibonacci, that one of them governs the entire activity of social man, and therefore that the mathematical basis of man’s sociocultural progress and of other natural growth systems is the same.
The level of aggregate stock prices is not a m _ tiqsity but a direct and immediate measure of the popular valuation of man’s total productive capability. That this valuation has aform is a fact of profound implications that should ultimately revolutionize the social sciences.
ENDNOTE: THE MENTATIONAL CONNECTION
It is also possible to link Fibonacci-based robust fractals in biology to a Fibonacci-based unconscious human mentation that governs impulsive herding behavior. This link completes a tentative explanation of how the Wave Principle is produced. For an introduction to this subject, please see the companion report, “Science v Is Revealing the Mechanism of the Wave Principle.”
Fractal objects whose properties are not restricted display selfsimilarity, while those that develop in a direction such as price graphs display selfafjnity. The term “self-similar” is often employed more generally to convey both ideas.
For more on this topic, see Johansen, A. (1997, December). “Discrete scale invariance and other cooperative phenomena in spatially extended systems with threshold dynamics” (Ph.D. Thesis). Somette, D. (1998). “Discrete scale invariance and complex dimensions.” Physics Reports 297, pp. 239-270.
Mandelbrot, B. (1988). The fractal geometry of nature. New York: W.H. Freeman.
Mandelbrot, B. (1962). Sur certains prix speculatifs: faits empiriques et modele base sur les processus stables additifs de Paul Levy. Comptes Rendus (Paris): 254, 39683970. And (1963). The variation of certain speculative prices. oumal of Business: 36,394419. Reprinted in Cootner 1964: 29 i -337. University of Chicago Press.
Mandelbrot, B. (1999, February.) “A multifractal walk down Wall Street.” Scientific American, pp. 70-73.
Gleick, J. (1985, December 29). “Unexpected order in chaos.” This World.
Sornette, D., Johansen, A., and Bouchaud, J.P. (1996). “Stock market crashes, precursors and replicas.” Journal de Physique I France6, No.1, pp. 167-175.
Mandelbrot, B. (1999, February.) “A multifractal walk down Wall Street.” Scientific American, pp. 70-73.
Elliott, R.N. (1938). The waue pinciple. Republished: (1994). RN. Elliott’s Masterworks - The Definitive Collection. Prechter, Jr., Robert Rougelot. (Ed.). Gainesville, GA: New Classics Library.
Elliott, R.N. (1940, October 1). “The basis of the wave principle.” Republished: (1994). RN. Elliott’s Master-works - The Definitive Collection. Prechter, Jr., Robert Rougelot. (Ed.). Gainesville, GA New Classics Library.
1 Arneodo, A., Argoul, R. Bacry, E., Muzy, J.F. and Tabbard, M. (1993). “Fibonacci sequences in diffusion-limited aggregation.” Growth Patterns in Physical Sciences and Biology, edited by Juan Manuel Garcia-Ruiz, Enrique Louis, Paul Meakin and Leonard M. Sander. New York: Plenum Press.
Weibel, E.R. (1962). “Architecture of the human lung.” Science, No. 137 and (1963) Morphometry of the human lung. Academic Press.
West, BJ. and Goldberger, AL. (1987, Jul/Aug). “Physiology in fractal dimensions.” American Scientist, Vol. 75.
Penrose, R. (1994). Shadows of the mind - a search for the missing science of consciousness. Oxford University Press.
Stanley, H.E., Buldyrev, S.V., Caserta, F., Daccord, G., Eldred, W., Goldberger, A., Hausman, R.E., Havlin, S., Larralde, H., Nittmann, J., Peng, CK, Sciortino, F., Simons, M., Trunfio, P., and Weiss, G.H. (1993). “Fractal landscapes in physics and biology.” Growth patterns in physical sciences and bioloa. Sew York: Plenum Press.
Arneodo, A., et al. (1993). “Fibonacci sequences in diffusionlimited aggregation.” Growth patterns in phyical sciences and biology.
Clouds and mountains, which are indefinite fractals, have a Hurst exponent near 0.8. Neurons (which grow as branching fractals) and the stock market (which grows as waves) have a Hurst exponent related to phi. These studies prompt me to suggest the hypothesis that fractal objects that manifest as branches or waves, i.e., the fractal objects of growth and expansion, will have a Hurst exponent related to phi, setting them apart from other fractal objects, which will have other Hurst exponents. What this means is that robust fractal objects split the difference between two Euclidean dimensions by .618, while other fractal objects do not. In other words, PhCrelated dimensionality is a property only of robust fractals.
Robert Prechter first heard of the Wave Principle in the late 1960s while studying psychology at Yale. In 1976, while at Merrill Lynch in New York, Bob began publishing studies on the Wave Principle. In 1978, co-authored, with AJ. Frost, Elliott Wave Principle-?@ To Market Behavior, and in 1979, started The Elliott Wave Theorist, a publication devoted to analysis of the U.S. financial markets. In November 1997, Bob addressed the International Conference on the Unity of the Sciences (ICUS) in Washington, DC, an international forum on interdiscipiinary scientific issues. The paper he presented at that conference was later expanded into his most recent book, The Wave Principle of Human Social Behavior and the New Science of Socionomics, which was published in 1999.
Return to Table of Contents
3: THE INTERACTION OF TRENDINESS MEASURES AND TECHNICAL INDICATORS
Basil Panas, CFA, CPA, CMT
PART I. INTRODUCTION
Background. Technical analysis has produced a plethora of indicators. Textbooks often classify them according to computational input: price, time, volume and sentiment. Practitioners need a taxonomy, which relates indicators to market phases. The art of technical analysis involves matching indicators with changing market conditions. Prices go through periods of trending and non trending. The implications of this are profound. Investors and traders must distinguish between trending and trading markets and adjust their trading tactics accordingly.
Definition of Trendiness. Although they are related, trendiness and volatility are different phenomena. A trend exists when prices are making higher highs and higher lows (uptrend) or lower highs and lower lows (downtrend). Trend is thus a function of the directionality of price changes. Volatility is a function of the size of price changes. Thus a strongly trending market displays both trendiness and volatility. However, a wide trading range displays little trendiness but much volatility. Finally, a very tight trading range is an example of low trendiness and low volatility. If market participants are to rely on different indicators depending on the trendiness of the market, they need to measure the directionality of price changes.
Hypothesis. This work proposes coordinating trend-following and counter-trend indicators using a measure of trendiness. The measure would characterize price action as trending or non-trending and thus select a trend-following or counter-trend indicator. The danger is that multiple indicators dilute each other’s effectiveness. The promise is that they become synergistic complements.
Theoretical Model. A simple regime-switching model was used to test the hypothesis. The model addressed three issues: how to trade in trending markets, how to trade in non-trending markets and how to distinguish between the two. Success depended on harmonizing the solution’s components.
The model employed exponential moving averages (EMA) for trending and Welles Wilder’s Relative Strength Index (RSI) (see bibliography) for non-trending markets. The Directional Relative Volatility Index (DRVI) described by Robert M. Barnes (see bibliography) measured the market’s trendiness or directionality and dictated whether EMA or RSI signals were taken. Testing Methodology. The test subjects were the 30 stocks listed in the appendix. They consisted of daily prices over various fiveyear periods. The stocks were divided into three groups according to their characteristic price action: trending, non-trending and mixed.
The benchmark tests consisted of EMAs and the RSI over lookback periods of 10, 20, 30 and 40 days. The hypothesis tests included these two indicators and the DRVI. The DRVI’s look-back period was 20 days. Its trendiness threshold was 0.5.
The tests were averaged for evaluation purposes. The limited number of parameters avoided the dangers of overoptimization. The tests assumed starting capital of $10,000 and commissions of $30 per trade which was executed at the next day’s opening price. All available capital was committed to each trade.
PART II. BENCHMARK TESTS
Background. Benchmark tests for all thirty stocks were established separately for the EMA and the RSI. These tests did not include a trendiness measure.
Exponentially Smoothed Moving Average
The trading rules for the EMA were:
Go long when: today’s close > EMA. Go short when: today’s close < EMA.
Table 1 summarizes the results. It gives the average return from all the EMA tests for each class of stocks. System close drawdown is the largest equity dip (relative to the initial investment) based on closed out positions. It is the maximum amount a closed out position fell below the initial investment amount.
As might be expected, the EMA performed best on trending stocks, worst on non-trending stocks and somewhere in between on mixed stocks. This was true of both measures of performance.
Relative Strength Index (RSI)
The trading rules for the RSI were as follows:
Go long when RSI crosses above 30. Stay long if RSI drops below 30.
Go short when RSI crosses below 70. Stay short if RSI drops below 30.
Table 2 summarizes the results. It gives the average return from all the RSI tests for each class of stocks.
As might be expected, the RSI performed best on non-trending stocks, worst on trending stocks and somewhere in between on mixed stocks. This was true of both measures of performance.
PART 3. TESTOF HYPOTHESIS Directional Relative Volatility Index
The DRVI scores trendiness from 0.0 to 1.0. The tests assumed a trend (trading range) when readings equal or exceed (are less than) 0.5. This threshold was used because it is the midpoint of the range. Empirical testing showed it was effective in separating trending from non-trending periods.
The trading rules for this system were of these:
The combination of an EMA, DRVI and RSI performed best on trending stocks, worst on non-trending stocks and somewhere in between on mixed stocks. This was true of both measures of performance.
Analysis of All Test Results
Table 4 compares the two sets of benchmark tests (EMA and RSI) with the tests of the composite (EMA, DRVI, and RSI).
The data suggest that combining a trendiness measure with technical indicators improves performance in certain cases. Regardless of the type of price action, trending, non-trending or mixed, better results were achieved with the composite model than the EMA alone. However, in the case of the RSI, the composite improved performance only in trending markets.
The implication is clear. A trendiness measure works best to eliminate whipsaw signals. This is consistent with the fact that whipsaws are usually associated with trend-following indicators (such as an EMA).
A visual inspection of the charts with their trading signals confirms this. Many bad EMA signals were eliminated by the DRVI. The DRVI did not, however, eliminate many bad RSI signals. Apparently, the RSI formula is better able to pinpoint the boundaries of a trading range than the DRVI.
The RSI compares prices to their own recent history while the DRVI compares readings to a threshold, in this case 0.5. Manipulating the DRVI trendiness threshold does improve results. Tests show that lowering the threshold in a trending market (to 0.25) makes the EMA more effective. This generates signals earlier in the trend. Raising the threshold in a trading range (to 0.75) eliminates more bad EMA signals and permits more accurate RSI signals. The problem is identifying trending and trading periods ahead of time.
The data do not show any predictive value in the DRVI trendiness readings. In fact, the DRVI signals changes in trendiness on a slightly lagging basis. This can be adjusted, as described above, by manipulating the threshold level.
PART IV. TRADING APPLICATIONS
Traders should filter the signals from trend-following indicators with a trendiness measure. They can enhance the measure’s effectiveness through its sensitivity setting or threshold. Traders can use traditional technical tools to identify trending and nontrending periods and adjust the threshold accordingly. For example, traders would use a high threshold as long as prices remain in a trading range. After a breakout (in either direction), they would switch to a low threshold. In uncertain markets they would default to a middle threshold.
I Achelis, Steven B., Technical Analvsis From A to Z, Chicago, IL: Irwin Professional Publishing, 1995.
Barnes, Robert M., Trading in Chonnv Markets, Chicago, IL: Irwin Professional Publishing, 1997.
Wilder, J. Welles, New Concerns In Technical Tradincr Svstems, Greensboro, NC: Trend Research, 1978.
Basil Panas earned a bachelor’s degree in Accounting from Rhodes University, South Africa. He is a CPA and holds the CFA designation. He has seven years of experience managing a fixed income portfolio ($60 million) for the City of West Covina, California, using both fundamental and technical tools. He is currently employed by the Metropolitan Transportation Authority in Los Angeles. He may be reached at 909/931-4926 or bpanas @ ibm.net.
Return to Table of Contents
4:HEAD AND SHOULDERS ACCURACIES AND HOW TO TRADE THEM
The Head-and-Shoulders pattern is probably one of the bestknown and venerable of chart formations. It is considered as one of the most reliable “by all odds” according to Edwards and Magee’s own words in their reference work.
Martin J. Pring quotes the Head-and-Shoulders as “probably the most reliable of all chart patterns,” while John J. Murphy’s analysis is almost identical when considering “probably the best known and most reliable of all major reversal patterns.” Some official legitimacy was gained in August 1995, when the New York Federal Reserve astonished both economists and technicians in publishing a computer study on the validity of the case: “Head-and-Shoulders: Not just a flaky pattern.” The old formation undoubtedly stands the test of time and represents a powerful tool in today’s trading and analysis. The suggestion is to invite you on a journey inside the Head-and-Shoulders. Some discoveries are still to be made. Rounding Bottoms and Complex Head-and-Shoulders are Multiple formations as well, and should be traded on a level of confidence that any technician should gain before acting. Traders have always been faced with some weakness when trying to profit from the pattern. It is not being irreverent to state that technical literature does not provide enough clear statistical accuracies on the subject. Most observations are pertinent orjudicious, but they hardly help when dealing with a trade to do or to avoid.
This paper will first specify what can be considered as a valid or adequate Head-and-Shoulders. Harmony limits and rules to follow will be shown according to classical practice. Secondle the study will analvze ‘known facts’ about Head-and-Shoulders. Probabilities and numbers will be put forward on the major topics such as Volume, Measuring Objective, Pullback and Pattern Length, among others.
In the third nlace. the naner will suggest trading techniaues to profit from the nattern and how to estimate the obiective. The entry level, the stop and three different ways of measuring the objective will be discussed. A complete track record will be established, showing the pattern degree of efftciency and the level of risk to take in order to make a living from it. Precise net valuations will be displayed.
Daily data from January 1990 till October 1997 have been selected on the S&P 500, US Treasury Bonds, Swiss Franc and Gold in an attempt to cover the major market sectors. Data are on a cash or spot basis in order to avoid roll-over gaps implied by the future market’s positive or negative carry.
The idea is to detect possible divergences between stocks, interest rates, currencies and commodities. Do Head-and-Shoulders develop the same way on various markets? Are all markets profitable? Is the Risk-Reward indisputable? These questions need ‘tentative’ precise answers.
Subjectivity is clearly the main difftculty when dealing with a pattern formation. ‘After the fact’ recognitions make trades more attractive than they are in the real world. Furthermore, patterns found on a chart may vary from one technician to another. Also, the picture may sometimes even prove to be rather different the following day for oneself!
However, well-trained individuals know very well that there is no room for various methods of assessment in that field; the margin is in fact pretty narrow. Despite the lack of statistics, many examples of Head-and-Shoulders are to be found in technical books, therefore diminishing misinterpretation. This work represents a full coverage of nearly eight years on four major liquid markets. All patterns discussed have been carefully selected in respect of the classical methodology as well as strict rules. The information and opinions contained have been compiled in good faith.
The author asserts that ethical standards of professional conduct have been highly respected. He is available, upon request, to defend any position taken or decision made.
This studv is based on dailv charts and deals solelv with Head-and-Shoulders which are tradable bv evervbodv, in contrast with patterns which are only caught by floor traders. The natterns selected in this studv meet two criteria. Thev are followed bv a Close bevond the Neckline, and a Pullback either to the Breakout level or the Neckline. on a dailv chart. In practice, you will have the time to analyze many markets and detect which one has just experienced a Breakout of a Head-and-Shoulders. Your next-day limit order will be easily calculated as well as your exit levels (Stop and Objective).
According to Robert D. Edwards and John Magee, the only qualification on an up-sloping Neckline is that the Bottom of the recesion between the Head and Right Shoulder must form appreciably below the general level of the top of the Left Shoulder. The logic applies for a down-sloping Neckline as well. In modern trading, the adverb ‘appreciably’ tends to disappear as commodities charts look more stretched than stock’s charts in the 1950s.
Multiple or Complex Head-and-Shoulders consisting of some Left and Right Shoulders, or even Tvvo-Headed, are common. However, the Neckline is not always easy to draw as two or even several possibilities often exist. Traders should take a position on any Pullback following an obvious Breakout, even in the case of multiple formations.
Head-and-Shoulders Tops or Bottoms are to be found at the end of a trend. It is not expected to consider as a true Head-and-Shoulders a pattern whose size is more than half the amplitude of the prior trend. John J. Murphy states that Reversal patterns can only be expected to reverse or retrace what has gone before them. In other words, the maximum objective is the size of the prior move. Therefore, too-big patterns may not reach their measuring objective, and imply a doubtful Risk-Reward ratio. Traders may avoid such trades which usually oblige them to place a Stop too far for fear of premature exit.
Symmetry is the key word for a Head-and-Shoulders pattern, even more so in a group of related formations which carry the same technical implications. Rounding Tops and Bottoms are Multiple formations as well and should be traded on a Pullback as soon as an obvious Breakout is detected.
The pattern has to be in Harmony with the environment. The word is somewhat romantic, but describes the kind of level of confidence any trader should gain before acting. Some technicians may consider this Gold-Comex development as a valid Head-and-Shoulders Top, but the assumed Right Shoulder represents, in fact, the move which negated the formation. The objective completion, a few days later, does not alter the picture.
Some situations are surprisingly not tradable. The Swiss Franc IMM picture looked promising in May 1997. Why was this trade not possible? This is a good example of an ‘After the fact’ trade. Whenever the first Breakout occurred, the Symmetry or Harmony of the Pattern was rather poor. The downward-sloping Neckline was steep, but not eliminatory. However, the Head and Left Shoulder distance as compared to the Head and Right Shoulder distance was a matter of worry. The lack of Harmony did not encourage the desire to trade when the Pullback eventually occurred.
Later on, the pattern looked more balanced and the decision to trade was logically taken. This time, the market did not give another chance to get in. We must learn to live with it.
One hundred and twenty one Head-and-Shoulders patterns have been found using daily charts on the S&P 500, Swiss Franc, US TBonds and Gold from January 1990 till October 1997 (94 months). Sixty percent of all formations were Head-and-Shoulders Bottoms, Gold recording an anomalous 76% of Bottoming patterns. Excluding Gold, Bottoming formations accounted for 53% of all patterns.
Sixty five of the 121 Head-and-Shoulders found experienced a Pullback powerful enough to initiate a trade. The entry or limit order has been placed at the Breakout point or the Neckline level, whichever was the less ‘ambitious.’ Whenever a Breakaway Gap occurred, the limit was placed at the less ambitious side of it (market should try to fill the Gap but may not succeed in true Breakaway situations).
Pullbacks have been seen 59% of the time in the case of Top formations, but 69% of the time in Bottoming ones. This is a prob able confirmation of the ‘gravity’ factor, showing that a market advance takes usually more time to develop than a market decline. Eighty percent of S&P 500 and US T-Bonds Bottom patterns experienced a Pullback after the Breakout. This analysis is not signi!ticant for currencies (either bullish or bearish, depending on the country). Gold had too few Top patterns to rely on the Pullback ratio observed in Top cases.
PULLBACK LIMIT ORDER
The classical method to determine the minimum Objective is based on the height of the pattern. The vertical distance from the Head to the Neckline is projected from the point where the Neckline is broken (purists use a logarithmic scale).
Our sample of 79 Head-and-Shoulders demonstrates that another method should be considered when trading patterns which are experiencing a Pullback (in other words, patterns which are tradable).
The minimum Objective should be estimated by measuring the vertical distance from the Bottom of the hollow between the Head and the Right Shoulder, up to the trendline joining up the Head’s crest and the Bight Shoulder’s crest.
That distance is then projected as in the classical style. An even more conservative method has been tested, using the vertical distance from the Right Shoulder’s crest to the h’eckline, then projected as in the classical style.
Cumulative total profits, on all trades, for the third method is 71.9%, clearly behind the classical measurement (77.8%) and the recommended one (79%)
Nearly 60% of minimum objectives have been reached using the recommended style, outperforming clearly the classical one (only 46% of objectives met). The conservative method managed to record an impressive, but misleading, 70% of success. This is where the Profit & Loss per trade comes into play, or the other side of the picture.
The performance per trade is unsurprisingly favorable to the classical method, almost compensating for the poor rate of success. However, method two is the best combination taking into account all the parameters.
The ‘Quality’ difference between method one and two is not so clear in terms of cumulative performances, but the picture is much better if we analyze the length of each transaction. On average, a trade (from the entry to the exit day) lasts 8 trading days with the recommended method, while it takes more than 10 days for the classical one. Considering that 2/3 of the trades in method 1 and 2 are identical, we have to understand what happens with l/3 of them. Analysis shows that the classical measuring objective is often too ambitious and is therefore missed. Then, the short-term trend reverses, and either the Stop limit is activated or the position sold at the Objective when the initial trend resumes, much later on. A position lasts 6 trading days with the conservative method. The potential move is, however, chronically underestimated.
Objective Not Tradable
Thirty live percent of patterns did not experience a sufficient Pullback and have been considered as ‘not tradable.’ In almost 100% of the cases, the market reached the target quickly, sometimes the same day as the Breakout occurred. Two thirds of the ‘not tradable’ patterns lasted less than 10 days. It is highly prob able that Pullbacks occurred on intra- day charts for the majority of these formations.
Volume characteristics are considered of critical importance in assessing the validity of the pattern. Activity is normally high during the formation of the Left Shoulder and tends to be quite significant, but lighter, when the price is at the peak. Right Shoulder is usually accompanied by lower Volume, a typical warning of diminishing buying activity during a Head-and-Shoulders Top, or the end of the selling pressure in the case of a Head-and-Shoulders Bottom.
Confirmation is provided in ranking the Volume: 55% of Left Shoulders recorded the highest Volume as compared to 32% for Heads and 13% for Right Shoulders. Objectives reached or not, the numbers barely change. Future patterns failures are therefore not to be found using that statistic alone.
The lowest Volume was recorded on 61% of Right Shoulders, 30% of Heads and 9% of Left Shoulders. Numbers were almost identical for Objectives completed and Objectives missed, which is again not helpful in detecting which Head-and-Shoulder is going to fail.
Thirty eight percent of the mid (or Nr 2) Volume has been recorded on Heads, 32% on Left Shoulders and 30% on Right Shoulders. Thirty four percent of patterns represented the ideal Volume sequence: Left Shoulder and the highest Volume, Head and the mid Volume, Right Shoulder and the lowest Volume. A small 4% developed in the most unusual way, with an inversed Volume sequence.
Volume at Bottom
Theory indicates that the most important difference between Head-and-Shoulders Tops and Bottoms is the Volume. At Bottoms, the market requires a significant increase in Buying pressure, reflected in higher Volume on up moves. The rally from the Head should show an increase of activity, often exceeding the Volume generated during the up move following the Left Shoulder.
Thirteen percent of Right Shoulders recorded the highest Volume, 5% at Tops and 8% at Bottoms. In this particular situation, 75% of Head-and-Shoulders Tops missed the recommended Ob jective, while 80% of Bottom patterns succeeded. This is an indication that a high Right Shoulder Volume is not comforting at Tops, but not really detrimental at Bottoms.
The Left Shoulder recorded the lowest Volume in 9% of all cases, 1% in Top and 8% in Bottom formations. Objectives have been met in slightly more than 50% of the formations.
The specific Volume number is not of major importance to the Technician. However, it is often necessary to classify the Volume into one of three categories: High, Low, Average. Giving a mark to each category (1 point for High, 2 points for Average and 3 points for Low), the sample shows an extreme similarity to the ‘grading’ study described before.
Forty one percent of patterns recorded the highest Volume (as compared to the Head and the Right Shoulder) and also a high Volume in amplitude, the typical case.
Despite this ideal situation, the recommended objective has been met in only 63% of the cases, not a significant hedge over ‘non typical’ situations.
Twenty five percent of patterns recorded the lowest Volume on the Right Shoulder and a low Volume amplitude as well. This scenario produced an impressive 80% of accomplished Objectives, a remarkable performance.
Measuring Objectives is discussed in terms of height, but too few studies deal with classical Objectives durations. In our sample, Head-and-Shoulders lasted 30 trading days, on average, from the start of the pattern until completion. A pattern started whenever the move which was at the very beginning of the Left Shoulder, crossed the future Neckline. The end of the pattern was materialized by the Breakout. Trades, initiated on the Pullback day, lasted 8 days, or 27% of the pattern’s duration, on average. This is a good indication of the time required when trading a Head-and-Shoulders. The durations of trades were identical for both reached and missed recommended Objectives, which means that the Stop order method was efficient (see Trading).
One third of trades lasted less than 20% of patterns’ durations (for example, less than 6 days on a 30 days pattern). One half were shorter than 30% and two-thirds less than 40%.
However, the most significant observation lies in the 50% or less category where 86% of trades lasted, at the maximum, half the durations of the patterns. This is a nice probability to put forward whenever a measuring Objective is activated.
Bottoms are generally flatter and generally take more time to develop, as the market falls to the floor more quickly due to the gravity effect. It does require a much greater effort for the market to launch a new Bull trend.
This characteristic is clearly confirmed by the current study. Top patterns lasted 23 days on average, while Bottom ones had durations of 34 days, a 50% differential.
Trades were completed after 8.3 days for Bottom formations, slightly above Top ones (7.6 days), showing that velocity was quite similar after the Breakout. Accordingly, Bottom trades tended to be shorter (as a percentage of patterns’ durations).
However, the major outcome was that 86% of trades lasted, at the most, half the size of all the patterns found for both Head-and-Shoulders Tops and Bottoms. Symmetry is perfect knowing that 44% of trades lasted, at the most, onequarter of the duration of all patterns for both Top and Bottom formations.
Lateral and vertical movements are proportional to each other as suggested by some theories, a statement of the obvious.
A Head-and-Shoulders is not complete until the Neckline is decisively broken on a closing basis: The Breakout Day. A close beyond the Neckline not only completes the pattern, but also activates the minimum measuring Objective. A sharp increase in Volume is usual during the Break out, a factor not always dominant in a Head-and-Shoulders Top. Following a Breakout, the market runs and quickly peaks. In 85% of cases, the Breakout’s peak was reached the day of the Breakout (Day 1) or the following day (Day 2).
The average Breakout’s peak, or incursion level, reached three-eighths of the expected measured move. In other words, three-eighths of the Objectives were accomplished before the Pullbacks.
Start of 24h trading as well as a high liquidity explained the absence of Gaps (only 5%)) still numerous on stocks charts.
RISK VS REWARD
The average gross profit at the recommended Objective was 1.497 higher than the potential loss at the Stop (Exit) level.
Head-and-Shoulders are reversal patterns. Thus, 78% of trades were initiated against the Mid term Secondary trend.
As mentioned before, this study deals solely with Head-and-Shoulders which are tradable by everybody. One hundred and twenty one Head-and-Shoulders patterns have been detected using daily charts from 1990 until 1997. Pullbacks occurred 79 times, allowing in practice anyone to enter all 79 trades (see Frequency, Pullback & Method). The trades are first analyzed on a very straightforward basis showing yearly gross gains and losses on each market.
Eighty nine percent of traded Pullbacks reached both the Neckline and the Breakout level. However, a limit placed at the most ambitious level (see Pullback) would have proved to be costly despite an estimated 10% ‘entry level’ savings. The total profit would have been cut by as much as 20%. Sixty three percent of trades generated a profit. The average profit per trade was 2.29%, much higher than the average loss of 0.90%. No loss above 2.50% had been recorded and a small 3% of trades lost more than 2%. Half of the winning trades exceeded 2% gain and 1 out of 10 exceeded 3% gain.
In order to value the net return of our 79 trades in the real world, the following rules have been established: $100,000 was the initial cash put in the account. A contract position in any market never exceeded 2.5 times the account’s value, a reasonable leverage which boosted the performance. Margins never rose above 15% of the net equity and could have been multiplied by 3, with positions in all 4 markets, without causing any disturbance for the trading. Round turn Commission and Slippage were $80 per contract.
At an average pace of 10 trades per year and knowing that each trade lasted 8 days on average, the interesting feature was the consistent high level of cash in the account.
Therefore, nothing could be more justified than trading other technical patterns like Double Tops & Bottoms or Triangles using the same account equity and the same system. Short term traders may use 10 days of intraday tick by tick charts and trade roughly 100 times per year.
Edwards, Robert D. and Magee, John; Technical Analvsis of Stock Trends 6th Edition, 1992 -,
Pring, Martin J.; Technical Analvsis Explained, 3rd Edition, 1991
Murphy, John J.; Technical Analvsis of the Futures Markets, 1986
I Murphy, John J.; Intermarket Technical Analvsis, 1991
Shaleen, Kenneth H.; Volume and Open Interest, 1991
Shaleen, Kenneth H.; Technical Analvsis & Ootions Strategies, 1992
Chang, Kevin P.H. and Osler, Carol; Federal Reserve of New York. August 1995 Reoort, Head & Shoulders: Not Just a Flaky Pattern
Etzkorn, Mark; Futures Magazine, January 1996 article, Fed& Shoulders
Since 1998, Serge Laedermann has been a partner at GF Geneva Finance, Geneva, Switzerland, focusing on Private Banking. Fundamental analysis, as well as technical analysis, is used in making investment decisions. Technical analysis is his major tool and he is mainly a specialist in pattern recognition.
In the 1980s Mr. Laedermann was a floor trader and a technical analyst at Credit Suisse and, later on, Chief Economist and Analyst at Bank of New York - IMB, Geneva.
In 1987 Serge cofounded the Swiss Association of Market Technicians (SAMT) and is currently a member of IFTA. (see over)
Return to Table of Contents
5: VOLATILITY AND STRUCTURE: BUILDING BLOCKS OF CLASSICAL CHART PATTERN ANALYSIS
Daniel L. Chesler, CTA, CMT
Like many technicians, I began my study of technical analysis with classical bar chart patterns: trusty head-and-shoulders, triangles, wedges and so on. Though I still rely on chart patterns today, not all technicians share my respect for this fm of analysis. Some technicians criticize classical chart patterns as being dqendent on the imagination of the chartist rather than on objective rules. While perhaps “the essence of charting is subjective interpretation, “I what Ifind even more interesting is the widespread and unapologetic use of classical chart patterns among successful analyts and traders2 In fact, the question of whether chartists assume a reality that does not exist seems almost moot given classical charting’s longevity over the past century.
Yet the question remains why classical charting, a technique that up pears to involve more exceptions than rules, attracts such a loyal following among otherwise skeptical professionals. What do these analysts and traders actually see when they identify a “classical” chart pattern? The answer I be&e does not lie hidden in the minutiae of traditional chart pattern definitions. More likely the answer is found in a set of general conditions that ex@-ienced chartists recognize intuitively.
Traditional chart pattern definitions stress the uniqueness of individual chart pattern shapes. For instance, think of the many variations on, the “triangle” theme alone: symmetrical, ascending descending, wedge type, inverted, inverted with rising or descending hypotenuse, continuation, reversal, top, bottom, et al., each with its own time, pice, and volume subtltties. It is my belief that in ascribing this much significance to individual patterns, we also understate the common thread that binds all chart pat terns.
In the following discussion I will try to &scribe that common thread by breaking chart patterns into generic components and examining each in turn before assembling them into a single model. My goal is to suggest a more compact and user-friendly a@roach to classical chart pattern analysis by focussing on the common elements that appear to characterize classical chart patterns in general.
First, I will review the history of price charts along with the background and basic tenets of classical chart pattern analysis. While these may be tired subjects for many readers, they are worth revisiting as they reflect the conventional views that we seek to expand. I will also discuss the role that classical chart patterns play within the broader scope of market analysis. Some of the practical strengths and weaknesses of classical charting will also be covered.
Next, a simple conceptual model will be presented, which attempts to depict classical chart patterns in terms of two basic components: the volatility component and the structure component. Individually these observations will not constitute new or unique theory on the subject of bar chart patterns or price behavior. Taken together, however, they should help reduce the degree of separation between what is typically perceived as a diverse range of classical chart pattern definitions. Using recent examples from the US stock market, I will show how the model can be used to simplify pattern recognition and enhance the timing of chart pattern-based trading decisions.
Again, my goal is not to advance a particular view of chart pattern analysis into the realm of verifiable science. Rather, I hope to add a measure of order to what some technicians view as the ambiguous process of finding and trading classical chart patterns.
PRICE CHART PRIMER
The earliest use of price charts has been traced back to 17th century Japan where it is believed price charts were first used to record and analyze the movements of the Japanese rice market.’ The use of price charts in the United States, however, did not develop until the late 19th century. Prior to the widespread use of charts in the U.S., price and volume analysis was generally limited to what one could observe and memorize as live quotes ran across a mechanical ticker tape. This practice became known generally as “tape reading.”
In the late lBOOs, the number of active stocks was few. However, as this number increased, following the list of active stocks on the tape became more difficult. Summarization of the data into price charts was the inevitable result.
Thus, a price chart is merely a graphic record of price and volume activity over a length of time -a graphic ticker tape so to speak. In this context one can understand how price and volume relationships gleaned from the practice of tape reading ultimately shaped charting principles. As one technician aptly put it, “tape reading was just primordial technical analysis.“
The earliest charts used in Western technical analysis are believed to be point-and-figure charts and existed at least fifteen years before the advent of bar charts5 Point-and-figure charts differ sub stantially from bar charts in that they do not specifically record time and volume data. They are noted for their ability to highlight “consolidation” zones, which generally imply either accumulation or distribution activity. The subject of this paper, however, relates only to bar charts and bar chart patterns.
Bar charts, probably due to their ease of construction, have been the most popular form of price charts since their introduction in the late /ewebBOOS./eweb Each “bar” consists of a vertical line representing the range of prices traded over a defined period: an hour, a day, a week, a month, etc. Prices are plotted on the vertical axis and time on the horizontal. Bar charts often include a graph along the bottom of the chart depicting volume activity and in the futures markets the open interest. The vertical axis of a bar chart is generally plotted on either an arithmetic or logarithmic scale, with the arithmetic scale being the more popular form. A logarithmic scale shows equal percentage increments of price rather than equal absolute increments as with an arithmetic scale.
CLASSICAL CHART PATTERN EVOLUTION
The 1948 book Technical Analysis of Stock Trends, written by Robert D. Edwards and John Magee, is often referred to as “the bible of technical analysis.” It is considered by many to be the definitive reference source for information on classical chart patterns. However, Edwards and Magee attributed the credit for their ideas to the original research and theories of both Charles Henry Dow and Richard W. Schabacker.
Dow was a co-founder of the Dow-Jones & Co. financial news service and the first editor of The Wall StreetJournal. He created the original Dow Jones stock averages in the late 1800s and wrote a series of editorials in the Journal that analyzed the price movements of these averages. After his death in 1902, William Hamilton and Robert Rhea refined Dow’s ideas into what became known as “Dow Theory.”
Loosely defined Dow Theory is a method of analysis that utilizes specific price patterns to infer the direction of the market’s primary trend. If prices are making a succession of new highs, interrupted by shorter-term reactions which terminate above previous reaction lows, the trend is considered to be up. Conversely, a succession of new lows in price accompanied by lower highs on intervening rallies indicates a downtrend. Dow recognized that on all levels, from major swings down to day-today fluctuations, prices do not move in a straight line along their trend but rather in a pattern of “zigzags” or “waves.” This observation by Dow is significant to chart pattern analysis as it forms the basis of all classical chart patterns; combinations of “zigzag” or “wave” patterns make up the core of all classical chart pattern definitions. Other Dow Theory principles also underlie classical chart pattern analysis. These include Dow Theory “lines” which appear as a narrow range of price fluctuations, and indicate a period of stagnation in price where buying and selling forces are roughly equal. As Edwards and Magee noted, a degree of coincidence appears to exist between Dow Theory lines and what might otherwise be viewed as classical chart formations.’ Finally, the idea that volume tends to expand on price movements in the direction of the dominant trend is also a tenet of both Dow Theory and classical chart pattern analysis.
While Dow focused on the longer-term trends of business activity as reflected in the relationship between the closing prices of his averages, it was Schabacker who adapted these principles to bar charts ofindividual securities on a short to intermediate time frame. In 1930, while employed as the financial editor of Forbes magazine, Schabacker authored Stock Market Theo-yy and Practice, a reference work on the subject of the stock market and trading. He also pub lished a manual in 1932, Technical Analysis and Stock Market Profits, which expanded upon the principles introduced in his first book. It was primarily through these two texts that Schabacker pointed out the various bar chart patterns that were later discussed and popularized by Edwards and Magee. Thus Schabacker was the chief architect of the “classical” chart patterns we know today such as triangles, head-and-shoulders, et al. To reiterate, these patterns belong primarily to the area of technical theory related to the trading of individual securities.
There have been other significant contributors to the body of charting knowledge, notably Richard D. Wyckoff and Ralph Nelson Elliott. Though it would be inaccurate to label the work of either Wyckoff or Elliott as “classical charting” per se, some overlap does exist. For instance, like Charles Dow, both Wyckoff and Elliott sought to identify repeatable price patterns of a cyclical or rhythmic nature.8 Wyckoff and Elliott also viewed the relationship between price and volume similarly to Dow.
More recently formal research has been made into the area of classical chart patterns. While no definite conclusions regarding the efficacy of classical chart patterns have been reached, there have been some encouraging results. For example, a 1995 study by the NewYork Reserve Bank found that the head-and-shoulders chart pattern yielded “significant excess profits” in select currency markets Research by Alex Saitta, a technician at Salomon, has shown profitable trading results using standardized classical chart patterns in the Treasury Bond market.“’
CLASSICAL CHART PATTERN BASICS
Most charting methods, including classical charting, make use of implied psychological or behavioral motivations. For instance, “doubt” is the emotion usually associated with the early stages of a new trend. After a trend has matured, “greed’! or “fear” are thought to be the forces that compel traders to “chase” prices up or down even farther, culminating in a frenzied “climax” of buying or selling activity. l1 Elliott wave structures are believed to directly reflect a rhythm in nature that manifests itself in “crowd behavior,” and ultimately in the shape of market prices.‘* Classical chart patterns, such as head-and-shoulders, triangles and others, are thought to be indicative of “pool operators” or “inside interests” who intentionally manipulate the market in distinct phases referred to as accumulation, markup, distribution, and markdown.
Regardless of the underlying causes attributed to their formation, classical chart patterns rely chiefly on the interpretation of trendlines, geometric formations and price and volume relationships. The primary chart patterns that Schabacker pointed out in his first book, Stock Market Theory and Practice, included patterns of accumulation or “bottoming,” and patterns of distribution or “topping.” Collectively these patterns are known as “reversal” patterns as they tend to coincide with a reversal of the prior established trend. Schabacker also identified a second group of patterns as “intermediate” or “continuation” patterns that are found “inserted in the progress of an already originated move.“14 As their name implies these patterns suggest only a pause in activity followed by a continuation of the preceeding trend.
The fact that a chart pattern appears as either a reversal or a continuation pattern does not rule out plentiful exceptions. For instance, an “orthodox” head-and-shoulders reversal pattern may develop into a continuation pattern, or vice versa. Most of the literature on classical chart patterns concedes this flaw. What can be said with moderate certainty however is that when prices have been in a trend and suddenly stop advancing or stop declining, they are now “doing something else.“i5 That “something else” is almost always the start of a classical chart pattern of one form or another.
Over time and depending on which analyst or trader you consult, individual patterns within each category have gone through minor name changes and other slight revisions. For example, Schabacker originally identified “wedges” as a reversal pattern, while other technicians have accepted the wedge pattern as both a continuation and a reversal pattern. However the names and categoies of the basic “area” patterns, which exclude all one and two-bar formations such as “island” and “gap” patterns, as well as “spike” or “V” reversals, can be broadly summarized as follows:
Patterns such as complex head-and-shoulders, irregular tops and bottoms, simple or “naked” trendlines, horizontal support and resistance lines, trend channels and others are also very much part of chart pattern vernacular. For sake of brevity, however, the patterns listed above safely represent the majority of all classical chart patterns.
In addition to identifying specific pattern “shapes,” classical chart pattern analysis also incorporates an analysis of the relationship between price and volume. For example, a price “breakout” is believed to confirm a pattern’s validity if it is accompanied by increasing volume. In the case of top reversal formations, this requirement is sometimes relaxed. However, in general, most chart patterns tend to follow a sequence of high and/or irregular volume in the early to middle stages, with markedly declining volume in the late stages, just prior to prices “breaking out” beyond the boundaries of the pattern. There is as Schabacker explained, “. . .the tendency for volume to decline during the period of formation of a technical area pattern. This shrinkage in activity is especially conspicuous as the formation nears completion, just before a breakout occurs.“i6 Charts IA-1C demonstrate actual examples of this behavior.
Another feature of classical chart patterns is the implied price target. Following the confirmation of a pattern, which is normally signified by a price “breakout,” chartists believe that targets can be determined that indicate how far prices will either rise or decline. The standard procedure for determining a price target is to measure the horizontal width of the pattern, in points or dollars, and then add or subtract this value above or below the point at which prices decisively exit the pattern.
Between the generous ridicule hurled at charting by well known market commentatorsn and the often exaggerated claims made by overzealous char&, it is probably safe to assume that classical chart patterns are a misunderstood subject. I have even known experienced technicians who mistakenly view classical chart patterns as a kind of esoteric knowledge for divining the future direction of stock prices. In the following section I will utilize quotes from various sources to help clarify the role of classical chart patterns.
It must be understood that chart patterns were conceived primarily as a “timing” or “trading” technique used for individual trade selection. Though Schabacker did find chart patterns useful as indicators of the general market, he did not view them as a longterm investment or market forecasting strategy; for this he considered fundamentals the more important of the two approaches:
“Our study has been devoted chiefly to consideration of the technical factors affecting stock market fluctuations. We have previously seen that such factors work much more swiftly and profitably than do the fundamentals. The technical side of the market is of special importance for the short-swing stock market trader - he who tries to take his quick profit and run, and then renew his operation in some other issue where technical considerations suggest another movement is about to materialize.“‘*
“The technical approach to the market. ..is based upon factors which relate chiefly, or at least more directly, to the market itself, to the price movement which results from the constant interplay between those who want to buy...and those who want to sell.. .“
“In other words, the fundamental factors suggest what ought to happen in the market, while the technical factors suggest what is actually happening in the market. It is, therefore, the more important of the two angles for the trader.. .‘
Thus Schabacker emphasizes the point that “technical factors” are particularly well suited to serving the needs of traders, or those who operate on shorter time frames. For Schabacker this specifically meant the use of bar chart patterns as a means of highlighting accumulation and distribution activity in individual stocks for the purpose of providing buy and sell signals.
The notion of chart patterns as a tool of the “timer” is as accepted among knowledgeable observers today as it was by Schabacker seventy years ago. For example, &-hard Aschinger, Professor of Economics at the University of Fribourg, Switzerland, makes an indirect but a propos reference to the nature of charting in a 1988 Swiss Bank Corporation article as follows:
” ‘Speculators,’ . . are defined as basing their investment policy on the behavior of the market itself, using recent patterns to predict future trends. . . . In reality, many chartists would fall into this category. . . . The point is that ‘fundamentalists’ usually follow a longer-range investment strategy, whereas ‘speculators’ have a basically short-term orientation.”
Aschinger implies that speculators are more concerned with matters of timing than with long-range “strategy.” He also links the use of “recent patterns” with the objectives of “speculators” as an accomplished fact. These views echo Schabacker’s and support the idea that chart patterns represent a technique belonging chiefly to traders.
Peter Brandt, one of Commodity Corp.‘s most successful traders for many years, and a speaker at the 14th Annual MTA Seminar in Naples, Florida, claims to rely almost entirely on classical chart patterns for making trading decisions. Brandt explains his views on classical charting in a 1990 book interview as follows:
“Classical charting is . . . useful only to highlight a certain defined trading opportunity. It is vital to keep in mind that over 50 percent of chart formations fail to deliver profitable trades. This may be an indictment of classical charting as a forecasting tool, but not as a trading tool. Classical charting principles do not explain all the markets all the time . . . . I am just looking for market situations that meet certain guidelines”
Thus, Brandt discounts any directional inferences of classical chart patterns. He views classical chart patterns as useful for the purpose of identifying and organizing individual trading decisions rather than for the purpose of outright prediction. For Brandt, chart patterns serve as a sort of bookmark that enables trades to be made with reference to a particular set of price levels, risks and potential outcomes.
The notion that classical chart patterns do not serve as a means of prediction is not necessarily a new idea.23 In the following quote attributed to legendary trader Jesse Livermore, Livermore appears to counsel that it is best not to place directional significance in chart patterns:
“In a narrow market, when prices are not getting anywhere to speak of but move within a narrow range, there is no sense in trying to anticipate what the next big movement is going to be-up or down. The thing to do is to watch the market, read the tape to determine the limits of the get nowhere prices, and make up your mind that you will not take an interest until the price breaks through the limit in either direction.”
One can assume that the “limits of the get-nowhere prices” which Livermore speaks of correspond to the boundaries of a classical chart pattern of some type. More importantly, Livermore reserves judgement regarding the future direction of prices “until the price breaks through the limit.” Thus, Livermore suggests that the forecasting value of chart patterns is subordinate to their main role of cordoning off the conditions that precede certain trends.
If we accept the idea that classical chart patterns are at best mediocre forecasting tools, then it follows that the successful use of chart patterns is dependant on the occurrence of a sufficient number of sustained trends to offset an even greater number of “false” signals. In this context, classical chart patterns are by necessity allied with the technical trend-following philosophy, which states that once a trend begins it is likely to continue.
In sum, two main points emerge regarding the role of chart patterns. The first point is that chart patterns are intended chiefly as an aid to trading and speculation of individual securities, although other uses such as general market analysis are also possible. The other is that chart patterns are not particularly useful as a means of predicting the future direction of prices; waiting for a decisive “breakout” in order to confirm the validity of a chart pattern would be unnecessary otherwise.
STRENGTHS AND WEAKNESSES OF CHART PATTERNS
Perhaps the greatest strength of classical chart patterns is their ability to help us participate in price trends. As trader and analyst William Gann noted, ” . ..the big profits are made in the runs between accumulation and distribution.“2’ Classical chart patterns offer traders a viable means of capturing these “runsn by highlighting the behavior which normally precedes significant trends.
In addition to highlighting specific trading opportunities, chart patterns can also be used to control risk by forewarning us of trend reversals. It is believed among most technicians that price trends do not reverse immediately, but rather go through a period of gestation before reversing. These periods often coincide with the development of a classical chart pattern. Those who wish to control their open position risk may find chart patterns useful in these situations.
Another strength of classical chart patterns is that they delineate when and at what price to buy and sell through the use of trendlines and price target objectives. Once the boundaries of a potential formation have been decided upon and marked off, these boundaries correspond to specific price and time coordinates that can be used to form specific trading and risk control strategy.
On the weakness side of the balance sheet, chart patterns are notoriously subjective entities. Surpluses of chart pattern examples exist in books and manuals with no corresponding supply of fixed pattern definitions. Thus there exists no simple way of determining whether or not an actual classical chart pattern has been discovered.
Because all classical chart pattern definitions are essentially approximations, chart pattern analysis contains the potential for abuse by portraying the personal biases of the chartist rather than actual market indications. The implied directional significance attached to specific chart pattern names, such as “Bearish Wedge” or “Bullish Triangle,” may also interfere with the chartist’s objectivity. To the extent that certain chart pattern shapes are associated with specific directional outcomes, the risk of taking on a preconceived directional bias by the analyst or trader seems inevitable.
Correctly identifying classical chart patterns in time to act on the “breakout” is also problematic. To borrow from Dow Theory parlance, how can one tell in what section of the line they are in until it is all over, and thus perhaps too late to take a position? Conversely, if we act too soon and pre-empt a chart pattern breakout, the result may be a series of “false starts,” also known as “whipsaws.”
As mentioned earlier, the conceptual model separates chart pattern behavior into two components: the volatility component and the structure component. Both are equally significant and their order is presented arbitrarily. Below I have summarized the primary aims of the model:
To offset the lack of classical chart pattern specificity by providing a less subjective though still not entirely fixed criterion for identifying patterns.
To serve as a notional benchmark for distinguishing valid chart pattern behavior from other types of market behavior.
To minimize the risk of implied directional biases by excluding the use of traditional “bull”, “bear” or pattern “shape” nomenclature.
To enhance the timing of trading decisions by more narrowly defining the specific behavior that coincides with chart pattern breakouts.
THE VOLATILITY COMPONENT
In lay terms, volatility is a measurement that tells us to what extent prices are changing over time. A market moving up or down 15 or 20 points a day is more “volatile” than the same market moving up or down in 3 or 5 point increments. Volatility can also serve as a proxy ofunderlying market activity. Using the same three stock examples from earlier, Charts 2A-2C demonstrate how changes in volatility, as measured by the one period range (highest high minus lowest low over the course of one day), correspond positively with changes in volume over the same time period. This phenomenon is not unique to daily stock charts; it can be observed across virtually all markets and time frames.
While the relationship between changes in volatility and changes in volume is by no means an absolute one, it is robust enough tc help us understand the dynamics behind chart pattern develop ment and the volatility component of the model. For example, i we assume that for every transaction there is both a buyer and : seller, volume can be viewed as a measure of the gross supply am demand at any point in time for a given market. In the case of ou model, volatility has been substituted for volume as a means of gauging these changes in supply and demand.
We can thus begin to describe the development of a classica chart pattern in terms of volatility as follows: In the final stages o a price trend, and at the beginning of a so-called “classical” char pattern, the market is characterized by relatively high volatility am wide price swings. Next, a gradual process of declining volatilit begins, leading at last to an area of suspense that marks the “begin ning of the end” of the chart pattern’s development. This fina stage immediately prior to a breakout is marked by a relative absence of price volatility versus the earlier stages of the chart pattern’ development. The market has reached a relative standstill and i positioned at the “tripwire” of an imminent breakout. Chart 3 depicts a schematic of the idealized volatility component.
Various tools can be used to help us measure changes in volatil ity that might not otherwise be obvious through visual inspectior of the chart pattern alone. The standard deviations of closing prices or an average of daily high-low ranges are two approaches. How ever, I prefer to use Welles Wilder’s Average Directional Index (ADX) which is based on an average of excesses between period to-period ranges, and is smoother in comparison to raw measure, of volatility such as standard deviation. Although ADX is normall; thought of as a measure of trend strength, this does not preclude the use of ADX for our purposes. 26 Later I will show how to utilize the ADX indicator (14 period) to gauge the changes in relative volatility that occur during chart pattern development.
THE STRUCTURE COMPONENT
The structure component of the model is not intended as a blue print that tells us where we are within the structure and hence when we are likely to go next, such as with Elliott wave or seasonal trad ing patterns. Rather, the structure component represents an ide alized form of cyclic behavior unique to classical chart patterns ir general. It is an attempt at making that which is important abou classical chart pattern “shapes” interesting - and not vice versa.
Specifically, the structure component emphasizes the tendency of chart patterns to exhibit a series of well-defined and periodic time cycles. This can be observed in most chart patterns as a series of distinct turning points marked by prominent highs or lows occurring at regular - or very nearly regular - time intervals. One possible rational for this phenomenon is that cycle periodicity is susceptible to greater distortion from the effects of trends. Hence, cycle periodicity is noticeably more discernible in non-trending environments as represented by so-called “classical” chart patterns.
In contrast, traditional chart pattern definitions focus primarily on the variation in cycle amplitude - or the “height” aspect of market time cycles as measured in dollars or points - as a means of classifying and distinguishing individual chart patterns. Traditional definitions rely on the repeatability of specific chart pattern “shapes” as formed by the combination of various cycle amplitudes. The model however is based on the assumption that generic conditions, such as declining volatility and distinct periodicity, underlie most chart patterns regardless of their shape or their individual “classical” definition.
The structure component also incorporates the tendency of classical chart patterns to exhibit noticeably overlapping cycles or “waves.” Most chart patterns reveal this tendency by taking on a horizontal orientation along the length of the pattern. This aspect of structure highlights one of the most fundamental differences between price trends and chart patterns: During price trends cycles overlap minimally, and in the case of very strong trends cycles may not overlap at all. Chart 4 depicts the idealized structure compo nent of the model.
In this section I will present several examples of how the model components combine to facilitate chart pattern based trading decisions.
Chart 5A, a weekly chart of Adobe, shows that the stock rallied strongly from a low of about 15 dollars in mid 1998 to a high of about 75 dollars in late 1999. Note the characteristic cycle structure during this trending phase; there is almost no overlap between adjacent cycles except for a brief consolidation during the early part of 1999. However, starting in mid November 1999, Adobe begins to retrace some of its gains. Upon closer examination of the daily chart during this phase (Chart 5B) we see an overlapping cycle structure and a distinct 18-19 day cycle periodicity. Thus the action in Adobe satisfies the basic requirements of the structure component of the model. Rather than attempt to attribute various meanings to the “shape” of this pattern, we are simply looking for generic behavior that is consistent with the model. Yet we are not ready to trade this pattern until we can satisfy the requirement of the volatility component of the model. In Chart 5C, we can see how relatively higher volatility, as denoted by ADX levels between 30 and 50 during the final months of 1999, coincided with a] the ending stages of the prior up-trend and b] the beginning stages of the chart pattern’s development. Note also how decreasing volatility, as depicted by gradually declining ADX levels, marked the late and final stages of chart pattern development. It is common to see ADX levels decline into the sub20 level immediately prior to the completion of a chart pattern, just prior to a pattern “breakout,” as Adobe demonstrates in January 2000. By waiting for the market to indicate through a measurable decrease in relative volatility its readiness to breakout, and by ignoring the directional implications of specific chart pattern “shapes,” we do not find ourselves engaged in the tricky game of constantly anticipating the time of the breakout or its direction.
Chart 6A is a weekly chart of Ames Department Stores, showing prices in a steep downtrend from mid June through November 1999. During this time Ames lost about fifty-percent of its value. Note the rally attempt in November beginning from point X on the chart, and the slight pullback in December to point Y. At this stage, on the heels of a multi-month decline in prices, a chartist might normally be pondering whether this current pattern represents a “higher low” or some other popular formation indicative of the early stages of a reversal. However, since we are only concerned with whether and to what extent the pattern imitates the model, we do not refer to specific bull, bear or pattern “shapes.” A closer look at the daily chart (Chart 6B) shows that Ames has established a distinct 10-l 1 day cycle periodicity with clearly overlapping waves. Finally, in January of 2000, the stock breaks down through support near 25 dollars (Chart 6C). Note how this pattern breakout follows a decrease in relative volatility, as denoted by the ADX indicator declining into the sub20 level. Through an awareness of the conditions that precede pattern breakouts, we are less likely to enter a position based on a premature or “false” move outside of the pattern. We are waiting for the market to tell us when it is ready to move, rather than imputing our own biases to the pattern.
Lastly, Chart 7A shows a weekly chart of software maker Novell, with prices falling steadily from mid-July through October 1999. Not unlike in the previous example of Ames Department Stores, Novell loses roughly fifty-percent of its value over a multi-month period. Beginning in October, a period of consolidation occurs in which a distinct 1415 day cycle emerges (Chart 7B). By mid-December, ADX has declined to sub20 levels, a point at which we have normally come to expect a breakout (Chart 7C). Although I have highlighted the detail around this pattern to simulate a classically styled “complex” or “irregular” head-and-shoulders bottom reversal, this was done purely in hindsight. The point is that such interpretations are open to wide debate; no doubt many technicians could have found different “classical” patterns in the chart prior to the upside resolution of prices in Novell in December 1999.
Merely stating a technical observation does not elevate it to the status of eternal truth. Yet, distilling our observations into strict rules also has its drawbacks; fixed rules inevitably fail to address the exceptional cases. The conceptual model offers a possible middle ground. It attempts to remove some of the subjectivity involved in chart pattern analysis while still permitting flexibility. The model is useful, even if it is not always an absolute indicator, if it helps us to understand the nature of the relationship between trending and non-trending markets, and how changes in volatility reflect changes in overall supply and demand.
We have seen how higher volatility coincides with the early stages of chart pattern development and declining volatility with the later stages. This has a logical basis: A more active market attracts and supports more participants, and hence more gross supply and demand - or total investor interest - than does a less active market. Any sudden changes in supply or demand in a less active or “quiet” market can result in sharply higher or sharply lower quotes due to a sheer lack of available buyers or sellers, hence resulting in what we commonly refer to as a pattern “breakout.” In the case of the structure component, we have seen examples of how chart patterns, regardless of whether they be reversal or continuation patterns in “classical” terms, can be set apart from trends by their characteristic periodicity and wave structure.
If we accept the idea that classical chart patterns can be broadly characterized by general conditions, rather than by a variety of pattern “shapes,” then perhaps classical chart patterns are truly not the products of wishful or delusional thinking as some critics allege. Unlike UFOs, we can point to evidence that supports chart pattern existence in the form of the volatility and structure components. In addition, we can utilize this “template” view of chart pattern construction to help us locate and trade patterns without debating over myriad chart pattern definitions and their directional significance.
Don Dillistone responded to my request for background information on charting and pointed me towards specific resources in the MTA library; John McGinley also offered several suggestions. Bruce Kamich graciously provided copies of out of print material by D.G. Worden. Alan M. Newman provided copies of material by Gerhard Aschinger. Mike Moody offered help in verifying background information.
Fosback, Norman G. [ 19761. Stock Market Logic, Dearborn Financial Publishing, Inc., pp. 213214.
John Murphy, Louise Yamada, Alan Shaw, Justin Mamis, Ned Davis, Alex Saitta, Bruce Kamich, Ralph Bloch, William O’Neil, John Tirone, Peter Brandt - these names represent a sample of well known market analysts and traders who utilize classical chart patterns.
Shaw, Alan R. [ 19881. Technical Analysis - reprintedfrom Financial Analyst? Handbook, DowJones-Irwin, Inc., pp. 313.
Dines, James [ 19721. How the Average Investor Can Use Technical Analystifm Stock Profits, Dines Chart Corp., pp. 171. Dines was paraphrasing - Worden, D. G., [date unknown]. Article “Tape Reading in an Old and New Key” in the Encyclopedia of Stock Market Techniques, pp. 820.
Murphy, John J. [ 19861. Technical Analpis of the Futures Markets, New York Inst. of Finance, pp. 322-323.
Murphy, John J. . Technical Analysis of the Futures Markets, New Ymlz Inst. @Finance, pp. 322-323.
Edwards, Robert D. and Magee, John [ 19921. Technical Analysis of Stock Trends, John Magee Inc., pp.203.
Edwards, Franklin R. and Ma, Cindy W., . Futures and Options, McGraw-Hill, Inc., pp. 444.
Osler, C.L., and P.H. Kevin Chang . “Head-and-shoulders: Notjust a flaky pattern,” paper, Federal Reserve Bank of New York, August.
Saitta, Alex [ 1998 1. “Reversal Formations: Predictive Power?,” article, Technical Analysis of Stocks and Commodities, November.
Shaw, Alan R. [ 19881. Technical Analysis - repn’ntedfrom Financial Analyst’s Handbook, Dow Jones-Irwin, Inc., pp. 316-317.
Koy, Kevin . The Big Hitters, Intermarket Publishing Corp., Interview with Robert Prechter, pp.159.
Schabacker, Richard W. [ 19301. Stock Market Theory and Practice, B. C. Forbes Publishing Co., pp. 626.
Roth, Phil . Technica& Speaking, interview, Traders Press, Inc., pp. 346.
Schabacker, Richard W. [ 19301. Stock Market Themy and Practice, B. C. Forbes Publishing Co., pp. 601.
Schabacker, Richard W. [ 19321. Technical Analysis and Stock Market Profits, Pitman Publishing, pp. 296.
In just one example, the headline of Louis Rukeyser’s March 1997 newsletter declares: “Leaving History to the Elves, This Market’s Charting Its Own Course.” Rukeyser goes on to say in big, bold print “The typical elf lives in the demonstrably vain hope that even-short term market action is scientifically predictable, if only one can tweak the chart one more time.”
Schabacker, Richard M’. [ 19301. Stock Market Theoq and Practice, B. C. Forbes Publishing Co., pp, 658.
Schabacker, Richard W. [ 19341. Stock Murk&Pro&s, B. C. Forbes Publishing Co., pp. 101.
Schabacker, Richard W. [ 19341. Stock MurketPr@ts, B. C. Forbes Publishing Co., pp. 101.
Aschinger, Gerhard [ 19981. “Reflections on the Crash,” article in the Swiss Bank Corp. journal: Economic and Financial Prospects, August/September issue.
Brandt, Peter L. [ 19901. Trading Commodity Futures with Classical Chart Patterns, Advanced Trading Seminars, pp.1428.
In the literature of Schabacker, Wyckoff, Edwards and Magee, Jiller, Brandt et al., there is general consensus that bar chart patterns are at best fallible as forecasting tools. Schabacker connives to place above average confidence in the predictiveness of some chart patterns, but not without disclaimers such as: “ . . .accurate analysis depends on constant study, long experience and knowledge of all the fine points...” (Stock Market Profits, pp. 35.)
Lefevre, Edwin [ 19231, Reminiscences of a Stock Operator, John Wiley & Sons, Inc., pp. 125.
Gann, William D. [ 19231. The Truth of The Stock Tape, Financial Guardian Publishing Co., pp. 125.
In the book Martin Pring on Momentum, International Institute for Economic Research, [ 19931, pp. 200, Pring gives an explanation of how ADX can be used to indicate declining “directional movement” as a precursor to new market trends.
Brandt, Peter L. [ 19901. Trading Commodity Futures with Classical Chart Patterns, Advanced Trading Seminars
Dewey, Edward R. and Dakin, Edwin F., [ 19471. Cycles -The Set ence of Prediction, Henry Holt & Company, Inc.
Dice, Charles A. and Eiteman, Wilford J. [ 19411. The Stock Mur- ket, McGraw-Hill, Inc.
Dines, James [ 19721. How the Average Investor Can Use Technical Analysis fw Stock Profits, Dines Chart Corp.
Edwards, Franklin R. and Ma, Cindy W., . Futures and Options, McGraw-Hill, Inc.
Edwards, Robert D. and Magee, John [ 19921. Technical Analysis of Stock Trends, John Magee Inc.
For@, Randall W., [April 28,1995]. Fed Gets Technical, Barron’s, page MWIO, Dow Jones & Co., Inc.
Fosback, Norman G. [ 19761. Stock Market Logic, Dearborn Financial Publishing, Inc.
Foster, Orline D. . The Art of Tape Reading Ticker Technique, Investor’s Press, Inc. - 1965 ed.
Frost, John and Prechter, Robert. [ 19781 Elliott Wave Principle, New Classics Library, Inc.
Gann, William D. [ 19231. The Truth of Tke Stock Tape, Financial Guardian Publishing Co.
Hamilton, William P. [ 19221. The Stock Market Barameter, Harper & Brothers Publishers.
Hurst, J. M. [ 19701. The Pn@ Magic of Stock Transaction Timing, Prentice-Hall, Inc.
Jiller, William L. [ 19671. How Charts Can Help You in The Stock Market, Trendline.
Kaufman, Perry J. [ 19871. The New Commodity Trading Systems and Methods, John Wiley & Sons, Inc.
Koy, Kevin [ 19861. TheBigHittem, Intermarket Publishing Corp.
Lefevre, Edwin [ 19231. Reminiscences of a Stock @eratar, John Wiley & Sons, Inc.
Murphy, John J. [ 19861. Technical Analysis of the Futures Markets, New York Institute of Finance.
Nixon, John Brooks, Jr. [ 19581. The Seven Fat Years: Chnmtiles of Wall Street, Harper & Brothers Publishers.
Plummer, Tony [ 19891. The Psycho&y of Technical Analysis, Probus Publishing Company.
Pring, Martin J. [ 19931. Martin pring on Market Momentum, Institute for Economic Research, Inc.
Schabacker, Richard W. [ 19301. Stock Market Theary and Practice, B. C. Forbes Publishing Company.
Schabacker, Richard W. [ 19321. TechnicalAnalysis and Stock Market Pro@, Pitman Publishing - 1997 ed.
Schabacker, Richard W. [ 19341. Stock Ma&tPn$ts, B. C. Forbes Publishing Company.
Schultz, Harry D. . A Treasury of Wall Street Wisdom, Investor’s Press, Inc.
Shaw, Alan R. [ 19881. Technical Analysis - m@intedfiDm Financial Analyst’s Handbook Dow Jones-Irwin, Inc.
Sheimo, Michael - “Michael Sheimo on Dow Theory”- Technical Analvsis of Stocks & Commodities, June 1998.
Sklarew, Arthur [ 19801. Techniques of a Professional Commodity Chart Analyst, Commodity Research Bureau.
Sperandeo, Victor . Trader Vie - Methods of a Wall Street Master, John Wiley & Sons, Inc.
Pistolese, Clifford [ 19941. Using TechnicalAnalysis, McGraw-Hill, Inc.
Wilkinson, Chris [ 19971. Technically Speaking, Traders Press, Inc.
Wyckoff, Richard D. [ 19101. Studies in Tape Reading, Fraser Pub lishing Company, Wyckoff/Stock Market Institute, Course Units 1 and 2. SMI, Phoenix, AZ tel: (609)-942-5165.
After graduation from Babson College in 1988, Dan Chesler began his career as a cash commodity trader, buying and selling in diverse markets ranging from industrial tomato paste to wheat and corn. Dan joined the Louis Dreyfus Group of companies in 1992 as a price-risk manager where he helped manage the world’s largest citrus products hedging and arbitrage program. In 1996 he worked as an analyst and trading assistant for a medium sized, managed futures fund. Currently he is a partner in a Miami based proprietary trading firm. Dan lives near Palm Beach, Florida
Return to Table of Contents